2. Theory and Methods 26
2.8. Molecular Modeling and Visualization
Drug resistance of HIV-1 Protease against JE-2147: I47V mutation investigated by
Molecular Dynamics simulation
Drug resistance of HIV-1 Protease against JE-2147: I47V mutation investigated by Molecular Dynamics simulation.
3.1. Introduction:
HIV is a retrovirus and its genome is encoded by RNA, which is reverse-transcribed to viral DNA by the viral reverse transcriptase (RT) upon entering a new host cell. This is followed by integration, transcription, translation and finally assembly and budding of HIV-proteins. Drugs against HIV have been developed by targeting key players at different points of the viral life cycle. To date, Food and drug administration (FDA) has approved anti-HIV drugs targeting three different viral proteins, HIV-RT, HIV-protease (HIV-pr) and HIV-gp41. Nine FDA approved drugs target HIV-pr, a member of the aspartyl protease enzyme family, which functions at the late stage of infection by cleaving viral Gag and Gag-Pol polyproteins thereby generating mature infectious virions. The knowledge that incomplete processing of these virions produces immature, non-infectious viral particles lead to intense research to find HIV-pr inhibitors [Swain et al. 1990, Thaisrivongs et al. 1996, Wang et al. 2000, Perryman et al. 2004]. This is also one of the major success stories of developing drugs using structure-based techniques using a combination of x-ray crystallography and computational work [Wlodawer et al.1998].
The major problem associated with the existing drugs against HIV-pr is drug resistance of different kinds. The drugs against HIV-pr are losing their effectiveness due to rapid point mutations in the genome of HIV. Two kinds of mutations have been identified for HIV-pr. One kind occurs at or near the active site by affecting the binding of drugs directly by reducing the van der Waals (VDW) contacts, increasing steric hindrance and increasing the number of unfavorable electrostatic interactions between HIV-pr and inhibitors [Wittayanarakul et al. 2005]. The other kind occurs far from the active site with enhancement of the enzymatic function [Piana et al. 2002b] by conformational changes that increase the affinity of the protease for substrates over inhibitors, the so-called compensatory mutation. Computational studies have been extremely valuable in
investigating the molecular basis of drug resistance since from static X-ray structures alone it is difficult to understand the protein conformational change and dynamics.
In the current work we focus on HIV-pr by considering a very high resolution (1.09 Å) crystal structure of HIV-pr with ligand JE-2147 [Reiling et al. 2002]. This high resolution structure allows closer look into the drug resistance mechanism of HIV-protease. JE- 2147, shown in Figure-3.1, is a peptidomimetic protease inhibitor developed by the company Agouron (Pfizer). In vitro data suggest that JE-2147 can be more efficient than the currently marketed HIV-pr inhibitors/drugs [Yoshimura et. al. 1999, Ohtaka et al.
2003]. JE-2147 is also interesting because of its unique resistance profile. There are two major mutations, which reduce the efficacy of JE-2147, I84V and I47V. While I84V is common for other similar ligands, I47V seems to be specific for JE-2147.
Figure 3.1: Structure of JE-2147. Atoms H, C, O, N and S are shown in color white, gray, red, blue, and yellow respectively. P1, P2, P1’ and P2’ represent different substituents. Atoms C1, C2 , C3 and C4 were used to define structural parameters (defined in the text) for analysis.
We have investigated the effect of I47V mutation using molecular dynamics simulation for both apo and complexed protein (for both WT and mutant) to get insight to this complex process. The outcome of the simulations shows that the mobility of the ligand
differs on mutation and also the mobility of residues in the flap region differs especially for the apo form.
3.2. Computational Details:
All four molecular dynamics (MD) simulations were started using the high resolution crystal structure of HIV-pr with JE-2147 (pdb ID 1KZK). Leap module of AMBER 7 program package [Case et al. 2002] was used to prepare the system for the simulation.
The AMBER 99 force field was used for the simulation [Cornell et al. 1995]. Charges of JE-2147 were calculated using the RESP [Bayly et al. 1993] procedure at the Hartree- Fock level with 6-31G* basis set after minimizing the molecule at the AM1 semi- empirical level [Dewar et al. 1985]. All four systems were immersed in a water box of size containing more than 8000 water molecules. TIP3P [Jorgensen et al. 1983] model was used to represent the water molecules. The positive charges of protein-water system were neutralized by chloride ions.
For the simulations of the complexed protein, charged state of the ASP25’ (one of the two catalytic aspartates) was taken as protonated. This is in accord with the finding in reference [Wang et al. 1996b] that ASP-25’ was protonated. Detailed calculations [Wittayanarakul et al. 2005] for HIV-pr and Saquinavir, a ligand similar to JE-2147 also found that one of the ASPs is protonated. The electrostatic interactions were calculated with the particle mesh ewald method [Essman et al. 1995]. Constant temperature and pressure conditions in the simulation were achieved by coupling the system to a Berendsen’s thermostat and barostat [Berendsen et al. 1984]. Bonds involving the hydrogen atoms were constrained to their equilibrium position with the SHAKE algorithm. The whole system was minimized for 200 steps. Then the system was heated to 300 K over 20 ps with 1 fs time step. Subsequently, 180 ps MD run was performed for equilibration. The time step for MD simulation for the production run was 2 fs. The system was then run for 3 ns and the stability of the trajectories was carefully monitored and first 300 ps were removed from the analysis. Since there is no structure available for the I47V mutant, 1KZK structure was taken as a template and the Ile47 (and 47’) was replaced by Valine.
3.3. Results and Discussions:
There are a number of attempts to understand drug resistance of HIV-pr using computational techniques. They differ in the particular mutation and particular inhibitor studied. Also, the lengths of molecular dynamics simulation covered a wide range (600 ps in reference [Wittayanarakul et al. 2005] to 22 ns in reference [Perryman et al. 2004]).
Analysis of the molecular dynamics trajectories was also performed in a variety of ways.
The main purpose of the present work is to understand the direct and indirect effects of I47V mutation. The following analysis is done keeping that in mind. In particular, the opening of flaps and protein-ligand movements are considered in the analysis.
3.3.1. Stability of the trajectories:
The stability of the trajectories of all four simulations was monitored by plotting the rmsd values of the C atoms as shown in Figure 3.2. It can be seen that the rmsd of all the trajectories are similar from the starting structures during the course of the simulations with values around 1.2-1.6 Å ensuring stable trajectories.
Figure 3.2: RMSD values for the C atoms for the WT and mutant simulation for both apo and complexed forms.
3.3.2. Comparing the apo proteins: WT vs. Mutant
Figure 3.3 shows the difference of isotropic temperature (B) factor between mutant and WT for each residue. The difference of B factor can give idea about the structural fluctuation of different regions of WT and mutant. The maximum changes in B-factor between WT and mutant are for the residues in the flap elbows of the two chains (34-35, 37, 35’-41’), flaps of the two chains (48-51, 46’-54’), part of the cantilever region (65-70, 65’-68’) though there are several other regions where B-factor differs. Thus from the B- factors, fluctuation of several residues especially the ones at the flap tips, flap elbows and cantilever region seems different between WT and mutant. In the next section we investigate these in more detail.
Figure 3.3: Difference of B-factor values from molecular dynamics (MD) simulation for WT and mutant HIV-pr simulation of the apo protein (mutant B-factor–WT B-factor).
3.3.2.1. Local fluctuations:
In this section, local structural differences between WT and mutant are investigated.
Especially the flap movement is investigated in detail. It is known that flap dynamics affects both drug binding and enzyme catalysis of HIV-pr. Several mutations affect dynamics of flap. For instance, L90M, G48V, V82F/I84V mutations open the flap more in the mutant than the wild type. On the other hand, M46I mutation makes the flap more closed. Recently a very detailed work by McCammon and co-workers on V82F/I84V mutant used a variety of parameters to find the flap dynamics. We have used some of those to enquire the extent of flap motion.
3.3.2.2. Flap tip to active site distance:
The distance between the flap tip (C of Ile50(50’)) and catalytic aspartates (C of Asp25(25’)) was measured from the simulation and the distributions of the same are shown in Figure 3.4(a) and 3.4(b) for chain A and B respectively. As seen from the figure, the distance between flap tip and catalytic site in chain A is clearly different for WT and mutant. The mean of WT distribution is 14.6 Å and standard deviation is 1.0 Å.
For the mutant, the mean and standard deviation are 13.5 Å and 0.6 Å respectively. So, the mean of these two distributions differ by more then 1.0 Å and WT distribution covers wider values. However, for chain B, the distributions are almost overlapping. The simulation results suggest that for chain A, the average flap tip to active site distance is less in the case of mutant. The difference in motion of chain A and chain B of HIV-pr has been observed in other simulation also [Wittayanarakulet al. 2005].
Asp25-Ile50 distance / Angs
6 8 10 12 14 16 18 20
Frequency
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35
WT MUT
Asp25'-Ile50' distance / Angs
6 8 10 12 14 16 18 20
Frequency
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35
WT MUT
(a) (b)
Figure 3.4: (a) Distributions of Asp25–Ile50 distance for both wild type (WT) and mutant HIV-pr simulation of the apo protein; and (b) distributions of Asp25'–Ile50' distance for WT and mutant HIV-pr simulation of the apo protein.
3.3.2.3. Ile50-Ile149 distance:
The distance between C of Ile50- C of Ile149 is shown in Figure 3.5. This will give the distance between the flap tips in the two chains. This quantity will can shed light on the difference of flap motion between WT and mutant. It is clearly seen from the figure that for the mutant the distribution has one peak around 8 Å, while for the WT the main peak is distributed around 7 Å and two other peaks around 9 and 6 Å. The mean and standard deviation of WT distribution are 7.7 and 0.9 Å and that of mutant are 8.2 Å and 0.7 Å
respectively. While there is considerable overlap between the two distributions the distance between the flap tips are fluctuating more in the case of WT than that of the mutant.
Ile 5 0 -Ile 5 0 ' distan c e / A n g s
5 6 7 8 9 10
Frequence
0.0 0.1 0.2 0.3 0.4
W T M U T
Figure 3.5: Distributions of Ile50–Ile149 distance for WT and mutant HIV-pr simulation of the apo protein.
Phe53’(C)-Cys67’(C) distance was also monitored (not shown in the figure) since there are large differences in the B-factors in those residues. The distributions are found to be different with some overlap. Though the B-factor is quite different for 53’ and 67’, the distance between them is not drastically different for WT and mutant. This is because the WT B-factor is more for 53’ but less for 67’, so the balance between these makes 53’-67’
average distance not drastically different for WT and mutant. Thus both from differences in B-factor and analysis of several parameters, we can conclude that the results of simulations indicate the dynamic motion of WT and mutant are different in certain residues. In particular, the distances between residues in the flap elbow and flap tips are different in the case of mutant and WT.
3.3.3. Comparing the complexed form of the protein: WT vs. Mutant.
3.3.3.1. B-factor analysis:
The difference of temperature factors (B-factor) of the protein in its complexed form is shown in Figure 3.6(a). It can be seen that compared to the apo protein, the difference between WT and mutant has reduced for most of the residues. However, there are significant difference for the residues 36-37 (flap elbow of chain A), 48-52 (flap tips of
chain A), 66(66’)-69(69’) (part of the cantilever regions in both chains). Figure 3.6(b) shows a comparison between X-ray structure B-factor and calculated B-factor for the WT. It can be seen that the calculated B-factors are much higher than that of the X-ray structure although the general pattern is mostly similar with few exceptions. It is to be noted that for high-resolution crystal structures (such as the one used here) the calculated B-factors can be much higher than experimental one. This is because the high-resolution crystals are usually very well packed, which is difficult to achieve in simulation done in pure solvent.
R e s i d u e N u m b e r
0 5 0 1 0 0 1 5 0 2 0 0
B-factor difference/Ang2
- 3 0 - 2 0 - 1 0 0 1 0 2 0 3 0
(a)
R e s i d u e N u m b e r
0 5 0 1 0 0 1 5 0 2 0 0
B-factor/Angs2
0 1 0 2 0 3 0 4 0 5 0 6 0 7 0
W T W T _ e x p
(b)
Figure 3.6: (a) Difference of B-factors for wild type (WT) and mutant complexed HIV-pr (mutant B-factor–WT B-factor); (b) B-factor for the complexed HIV-pr (calculated and experimental results for the WT).
Analysis of Local fluctuations of inhibitor complexed WT and mutant:
Next several key local motions were investigated. Three quantities out of those are discussed in the following (a) Asp25(25’)-Ile50(50’) and Ile50-Ile50’ distances which indicate the movement of the flap and flap-active site movement, (b) Asp25(25’)- inhibitor distance, which will be an indicator of the protein-ligand motion, (c) dihedral angles of Ile-47’ and the P2’ moiety of the JE-2147, which will shed light on the orientation of the protein and inhibitor.
3.3.3.2. Asp25(25’)-Ile50(50’) distance and Ile50-Ile50’ distance:
The Asp25(25’)-Ile50(50’) distances were monitored during the course of the simulation to compare with observed differences found in the case of apo proteins. It was found that for chain A the distribution is much narrower compared to the apo protein. Moreover, the WT and mutant distributions have significant overlap. This indicates that in the ligand bound state, the distance between flap tips and the active site does not differ significantly on mutation for chain A. For chain B, there is more difference between the distributions compared to chain A. To know the relative motion of the flap tips Ile50-Ile50’ distance was also monitored. Again it is much narrower than the apo form, and the difference between WT and mutant is much smaller.
Asp25-Ile50 distance/Angs
6 8 10 12 14 16 18 20
Frequency
0.0 0.1 0.2 0.3 0.4 0.5
WT MUT
Asp124-Ile149 distance/Angs
6 8 10 12 14 16 18 20
Frequency
0.0 0.1 0.2 0.3 0.4 0.5
WT MUT
Figure 3.7: Distributions of Asp25-Ile50 and Asp124-Ile149 distances for both WT and mutant simulation for the complexed protein.
The results of this section indicate that though Ile-Ile distance is similar in the complex there is still difference in the Asp-Ile distance in chain B. These three distributions are shown in the following Figures 3.7 and 3.8.
Ile50-Ile149 distance / Angs
5 6 7 8 9 10
Frequency
0.0 0.1 0.2 0.3 0.4 0.5
WT MUT
Figure 3.8: Distributions of Ile50-Ile149 (flap-flap) distance for the WT and mutant for the complexed protein.
3.3.3.3. Protein-ligand distance:
In previous work [Piana et. al. 2002b], it has been shown that the displacement of the substrate to the active site of the protein is coupled with the complex motion of the entire protein. We have used the definition given by Rothlisberger group [Piana et. al. 2002b] to get the protein-ligand distance in a simple manner; this is the average distance between the C of two catalytic Asp (Asp25 and Asp25’) and the two carbons of the ligand (C1
and C2 in figure 2). The distributions of the average of these four distances are shown in Figure 3.9.
This shows that the distributions are essentially same for WT and mutant, which indicates that the ligand is bound strongly to the catalytic aspartates and mutation does not have any significant effect.
Protein-ligand distance /Angs
7.5 8.0 8.5 9.0 9.5 10.0 10.5
Frequency
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35
WT MUT
Figure 3.9: Distributions of protein–inhibitor distances (see text) for both wild type (WT) and mutant HIV-pr simulation.
3.3.3.4. Dihedral angle reflecting the orientation of protein and ligand:
Residue Ile47' in the WT HIV-pr interacts with the P2' site of the inhibitor in the crystal structure. The loss of such molecular interactions may be a key direct effect of mutation.
To explore this, two dihedral angles involving Cβ–Cγ1 (Cγ2) of Ile-47' (and Val47' for the mutant) and C3–C4 of JE-2147 were monitored by simulations to give an indication of relative orientation of the P2' group of the inhibitor with respect to the side chain of the 47' residue of HIV-pr.
Dihedral Angle / Degree
0 100 200 300 400
Frequency
0.0 0.1 0.2 0.3 0.4 0.5
WT MUT
Dihedral Angle / Degree
0 100 200 300 400
Frequency
0.0 0.1 0.2 0.3 0.4
WT MUT
(a) (b)
Figure 3.10: Distributions of dihedral angles used to obtain orientation of P2' moiety of the inhibitor relative to the 47' side chain for both WT and mutant. (a) CG1 and (b) CG2 dihedrals.
Distributions for the two angles (henceforth denoted by CG1 andCG2) indicated significant differences between the WT and mutant HIV-pr (Figure 3.10).
The dihedral angle measurements for mutant HIV-pr showed a greater fluctuation when compared with that of WT, including values both in the regions near 0° and 360°. The mean and Standard Deviation of the WT dihedral angle were 36.5° and 13.9°, respectively, for CG1 and 146.4° and 17.5° for CG2, respectively. The mean and Standard Deviation of the mutant dihedral angles were 134.9° and 138.1°, respectively, for CG1 and 237.6° and 72.2°, respectively, for CG2. Thus, the relative orientation of residue 47' changes more rapidly and with a wider range of angles relative to the P2' group in the case of mutant HIV-pr. This results from of the loss of predominantly hydrophobic interaction between the P2' group of JE-2147 and mutant HIV-pr which effects increased mobility for the inhibitor. Both WT and mutant HIV-pr were analyzed for movement of water molecules, and simulations revealed that only one water molecule came within the 3 Å from the center of the four atoms chosen to define the dihedral angle. Therefore, the effect of water in the direct interaction between 47' residue of HIV- pr and P2' of JE-2147 is predicted to be minimal.
3.3.4. Overall molecular analysis of I47V HIV-pr resistance to JE-2147
In this study, drug resistance of 147V HIV-pr against an experimental inhibitor, JE-2147, was investigated to understand the effect of this mutation at the molecular level. For this purpose, four different simulations involving apo and complexed proteins (with and without the 147V mutation) were performed representing all atoms of the protein and inhibitor with thousands of explicit water molecules. The quality of the simulations were ensured by the rmsd values and by comparing with the B-factor obtained from X-ray for the complexed form of the WT. The results show that for the apo protein, there is difference in dynamic motion of WT and mutant involving the residues in the flap elbow and flap tip regions. In the case of the complexed HIV-pr, the difference in dynamic motion between WT and mutant was less than that of the apo protein relative to the motion of residues in the flaps and flap elbows. The average protein–ligand distance is
not affected by mutation. The most distinct motion for the HIV-pr complex was the movement of the side chain of Val47' about the inhibitor and this was greater than for Ile47' in the WT. Hence, such results correlate to losing one –CH2 group in the side chain between mutant and WT proteins, and consequent decreased hydrophobic interactions of Val47' mutant HIV-pr for inhibitor binding in the case of JE-2147. Such findings suggest that a larger group at the P2' position JE-2147 might provide recovery of such binding properties. Furthermore, the Val47' mutation causes a change in the dynamics of the flaps and such disruptions are likely to play a role in the binding and hence to the resistance.
3.3.5. Comparison of I47V with other mutants:
A comparison of the 147V mutation with other studies focused on different mutations is important to further interpret the findings described in this study. In this regard, we have considered the mutations G48V (flap region) and I84V (active site) to compare with I47V. The G48V mutation has been found for the clinical inhibitor Saquinavir (SQV), whereas I84V mutation exists for all clinical inhibitors. It is known that the major direct effect of G48V mutation is increased steric effect between protein and inhibitor, whereas the I84V mutation affects the dynamical motion of the protein.
The highly conserved flap tips of HIV-pr I47–G48–G49–I50–G51–G52–F53 are extremely flexible. Nuclear magnetic resonance (NMR) studies by Torchia and colleagues [Ishima et al. 1999] have shown that residues 49–52 have rapid motion in
<10 ns time scale and fluctuation of F53 is likely coupled to the motion of the entire flap.
Such G48V simulation studies [Wittayanarakul et al. 2005] have determined that there is difference in rmsd for the residues in this region upon mutation. It is noted that these reported [Wittayanarakul et al. 2005] simulations were performed with complexed HIV- pr, whereas in the present study simulation were conducted with both apo and complexed HIV-pr. It is likely that simulation studies of apo forms of G48V mutation would show more differences in flap motion, and such a finding has been verified in a recent report [Hamelberg et al. 2005] wherein accelerated MD calculations showed decreased flap motion in the G48V (apo form) to be substantially different from the WT. In the present study, we conclude both direct effects (VDW and steric for residues 47 and 48,